-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathresearch.html
More file actions
162 lines (116 loc) · 7.65 KB
/
Copy pathresearch.html
File metadata and controls
162 lines (116 loc) · 7.65 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<meta name="viewport" content="width=device-width, initial-scale=1">
<meta name="description" content="">
<meta name="author" content="">
<title>Projects</title>
<!-- Bootstrap Core CSS -->
<link href="css/bootstrap.css" rel="stylesheet">
<link href="css/hover.css" rel="stylesheet" media="all">
</head>
<body style="margin: 20px 0; font-family:Open Sans, sans-serif;">
<nav class="navbar navbar-default navbar-fixed-top">
<div class="container-fluid">
<!-- Brand and toggle get grouped for better mobile display -->
<div class="navbar-header">
<button type="button" class="navbar-toggle collapsed" data-toggle="collapse" data-target="#bs-example-navbar-collapse-1" aria-expanded="false">
<span class="sr-only">Toggle navigation</span>
<span class="icon-bar"></span>
<span class="icon-bar"></span>
<span class="icon-bar"></span>
</button>
<a class="navbar-brand" href="#">Corby Rosset</a>
</div>
<!-- Collect the nav links, forms, and other content for toggling -->
<div class="collapse navbar-collapse" id="bs-example-navbar-collapse-1">
<ul class="nav navbar-nav">
<li><a href="projects.html">Projects</a></li>
<li><a href="research.html">Research</a></li>
<!-- <li><a href="stuff.html">Good Links</a></li> -->
<li class="dropdown">
<a href="#" class="dropdown-toggle" data-toggle="dropdown" role="button" aria-haspopup="true" aria-expanded="false">Photos <span class="caret"></span></a>
<ul class="dropdown-menu">
<li><a href="black_and_white.html">Black & White</a></li>
<li><a href="art.html">Art</a></li>
<li><a href="https://goo.gl/photos/eqSxwjWWtwjuq5hW8">Random Things</a></li>
</ul>
</li>
<li><a href="http://www.linkedin.com/in/corbyrosset">LinkedIn</a></li>
<li><a href="mailto:crosset2@jhu.edu">Email</a></li>
<li><a href="https://github.com/corbyrosset">Github</a></li>
</ul>
</div>
</div>
</nav>
<!-- Project -->
<div class="container">
<!-- Page Heading -->
<div class="row">
<div class="col-lg-12">
<h1 class="page-header"></small></h1>
</div>
</div>
<!-- /.row -->
<div class="row">
<div class="col-md-5">
<a href="">
<img class="img-responsive" src="images/large/ranking-schematic.png" alt="Ranking Stack" title="Ranking Stack for a typical search engine">
</a>
</div>
<div class="col-md-6">
<h3>Optimizing Query Evaluations using Reinforcement Learning for Web Search</h3>
<p> As part of my research at Microsoft Bing, we developed a novel algorithm to search a very large index for documents related to a query. Rather than scanning the index for matching terms using hand-crafted rules, we propose a reinforcement learning approach that seeks to optimize the tradeoff between finding important documents and computation time. Please see our preprint on Arxiv which was accepted to Sigir 2018 in July.
</p>
<p>
<a class="btn btn-primary hvr-sweep-to-left" href="https://arxiv.org/abs/1804.04410">Submission</a>
</p>
</div>
</div>
<hr>
<div class="row">
<div class="col-md-5">
<a href="">
<img class="img-responsive" src="images/large/tsne-entities.png" alt="t-sne embeddings of entities" title="t-SNE visualization of learned knowledge graph entity embeddings">
</a>
</div>
<div class="col-md-6">
<h3>Knowledge Base Completion with Embeddings of Graphs, Text, and Paths</h3>
<p> This is research I conducted as part of a Senior Honors Thesis for the Computer Science Department with the generous support of the Pistritto Fellowship. I worked on algorithms for learning entity and relation-specific embeddings of knowledge graphs, and different training approaches for the task of knowledge graph completion. The abstract is here: </p>
<p> Knowledge bases are an effective tool for structuring and accessing large amounts of multi-relational data, but they are often woefully incomplete, especially in broader domains. We consider the task of learning low dimensional embeddings for Knowledge Base Completion and make the following contributions: 1) a novel embedding model, ModelE-X, that uses few parameters yet outperforms many state-of-the-art, more complex algorithms, 2) the realization that the often-unreported metric of relation ranking yields valuable insights into algorithms' behavior and 3) we scrutinize macro vs. micro-averaging of ranking metrics and discuss which is a better indicator of generalizability.
</p>
<p>
<a class="btn btn-primary hvr-sweep-to-left" href="files/KBC_modelEX_final.pdf">Short Paper</a>
<a class="btn btn-primary hvr-sweep-to-left" href="files/thesis-knowledge-base.pdf">Thesis</a>
</p>
</div>
</div>
<hr>
<div class="row">
<div class="col-md-5">
<a href="">
<img class="img-responsive" src="images/large/multiview-rep-learning.png" alt="t-SNE visualization of learned phone embeddings" title="t-SNE visualization of learned phoneme embeddings for speech recognition; note that the labels were not used during training">
</a>
</div>
<div class="col-md-6">
<h3>Deep Canonically Correlated LSTMs</h3>
<p> This was a research project with my colleague Neil Mallinar for his Thesis at Johns Hopkins University. We wanted to investigate whether the CCA objective could be used in time series settings. The abstract is here:
</p>
<p>
We examine Deep Canonically Correlated LSTMs as a way to learn nonlinear transformations of variable length sequences and embed them into a correlated, fixed dimensional space. We use LSTMs to transform multi-view time-series data non-linearly while learning temporal relationships within the data. We then perform correlation analysis on the outputs of these neural networks to find a correlated subspace through which we get our final representation via projection. This work follows from previous work done on Deep Canonical Correlation (DCCA), in which deep feed-forward neural networks were used to learn nonlinear transformations of data while maximizing correlation.
</p>
<p>
<a class="btn btn-primary hvr-sweep-to-left" href="https://arxiv.org/abs/1801.05407">Short Paper</a>
</p>
</div>
</div>
</div>
<!-- /.container -->
<!-- jQuery -->
<script src="js/jquery-2.1.1.js"></script>
<!-- Bootstrap Core JavaScript -->
<script src="js/bootstrap.min.js"></script>
</body>
</html>